A reflection on speed, capability, and the cost of never stopping
The first time you use AI to build something and it just works -- actually works -- something shifts. A task that would've taken an afternoon wraps up in twenty minutes. You sit back and blink. That was it?
There's a rush to it. Not just the time saved, but the feeling of leverage. You described what you wanted and it appeared. You hit a wall, asked a question, and the wall moved. The distance between idea and implementation collapsed overnight.
For me it was a data pipeline I'd been putting off for weeks. The kind of refactor you know needs to happen but you keep deprioritizing because the tedium isn't worth the time investment. I opened Claude Code, described the problem, and ninety minutes later it was done. Not hacked together. Done well.
I remember thinking: if everything is this fast now, imagine what I could get done.
That thought felt like possibility. It was the first step into something else entirely.
It starts small. You finish a task early, so you pick up the next thing on the list. That goes fast too. So you grab another.
By the end of the day you've cleared a week's worth of work and you feel incredible. You're shipping features, writing docs, fixing bugs that have been sitting in the backlog for months. People notice.
"If that only took ten minutes, I can squeeze in one more thing before the meeting." "That was easier than I thought. Let me knock out this other one while I'm in the zone." "It's only 11pm. I could probably get this deployed tonight."The list never gets shorter. It regenerates. Every task you complete reveals two more you didn't know existed. And because each one feels so achievable now, saying "I'll do that tomorrow" feels almost lazy.
You have the tools. You have the momentum. Why would you stop?
Your calendar is full. Your IDE has nine tabs open. Slack is pinging. You're mid-deploy on one project while reviewing a PR on another while your phone buzzes with a Teams notification about a meeting that started three minutes ago.
The strange part is you're not struggling. You're keeping up. Every ball is in the air and you're catching all of them. The AI handles the boilerplate, you handle the decisions, and together you're moving at a pace that would've been physically impossible a year ago.
But there's a cost you're not counting.
You haven't taken a real break in weeks. Lunch is something you eat while reading documentation. Your evenings have become extensions of your workday because the work is genuinely interesting and the tools make it so easy to keep going.
"Just one more thing" has become your default state.The feeling of leverage has been quietly replaced by the feeling that stopping means falling behind. You're not burned out because you're failing. You're burning out because you're succeeding.
And then you realize you haven't looked away from a screen in eight hours.
Your neck hurts. Your eyes are dry. It's dark outside and you don't remember when that happened.
The thing you shipped at midnight could have waited until morning. The thing you shipped this morning could have waited until next week.
The tools didn't do this to you. You did this to you.
Here's what nobody tells you when they talk about AI making you more productive: your capacity hasn't actually changed. Your biology hasn't been upgraded alongside your toolkit. You still need sleep. You still need to look at trees and talk to people and eat meals that aren't at a keyboard.
What changed is the speed of the work, not the speed of you.
I've been watching people in the data community -- people like Kurt Buhler, who is one of the more thoughtful voices on this stuff -- talk openly about this same tension. The tools are extraordinary. The output is real. But so is the exhaustion that comes when you treat every saved hour as an hour you should now fill.
Efficiency was supposed to buy us time.
Somewhere along the way, we started spending it before it arrived.
The antidote isn't to stop using the tools. That ship has sailed, and honestly, the tools aren't the problem. The problem is the instinct to fill every gap. To optimize every minute. To feel like if you can do something, you should do it right now.
Finish the work faster -- and then close the laptop. Go for a walk. Let the backlog exist without guilt. The tasks will be there tomorrow, and you'll still be fast.
The most productive thing you can do with the time AI gives you back is nothing.
If this resonated, you're probably already doing too much. Take that as your sign to stop for today.
This page was built with Claude Code -- the same kind of tool this piece is about. The irony isn't lost on me. But it was built during working hours, and I closed my laptop when it was done.
Inspired by the surreal HTML blog writing of Kurt Buhler. KBHM